4 things you should know about your hard drive

PC-Net's Utility News

Things That Go Bump in the Drive--Part 1
(Or 4 Things You Should Know About Your Hard Drive)

By Don Watkins

February 1999 (last updated October, 2007)

This is the first of a two-part discussion of ways to verify that hardware is working properly. We'll also look at the tools used to verify overall disk integrity. Next month, we'll examine data recovery tools that can be used to recover information from an uncooperative hard drive. Naturally, the best recovery tool is a full, current backup, but that's not always an option.

1. Boot That Floppy

One of the first things you should have is a bootable floppy. Windows makes preparing one a simple task: Start | Control Panel | Add/Remove Programs | Startup Disk. Windows 98 users have it even easier, as the disk that is created includes drivers for most CD-ROM drives. Windows 95 users will need to obtain a driver from their CD-ROM drive vendor, copy MSCDEX.EXE from \Windows\Command and make the necessary entries in Config.sys and Autoexec.bat.

For Windows XP and Vista it's even easier. The distribution CD includes an option to install or run the Windows recovery console. The recovery console will give you the same command prompt that a boot floppy will plus the added benefit of having all the files and programs on the CD-ROM at your fingertips.

Sadly many systems don't come with a "pure" Windows CD but rather a "recovery" CD that only allows you the option of restoring files. In this case you'll want to use the boot floppy option.

2. Advise That Data

I once had a hard drive that was exhibiting bad behavior, giving clear signs that the drive itself had a problem. I contacted the vendor, who told me to download Ontrack Data Advisor and run it on the drive. The cost of the program is modest and the program prepares a self-booting floppy with its own operating system. It's a very simple procedure and is well worth the download time. After supplying the error code that Data Advisor reported, the vendor confirmed the drive was defective and replaced it under their warranty program. I recommend that you take a look at this program if you are interested in checking the physical health of your drive.

3. The Will to Fail

Steve Gibson, author of Spinrite and a computing giant, said it best: "There are two types of hard drives--those that have failed and those that will." I often wonder if Steve is a fellow Cubs fan; nevertheless, Spinrite is the best program I have used to work with hard drives. I won't attempt to describe the full level of detail this program can deal with. Spinrite has options to perform a surface scan of the drive, or to go deeper and examine the internal surfaces. It can stress test the drive and identify potential failing areas and also resurrect areas that were marked as bad by other programs like Scandisk. If you do work with hard drives, or have experienced problems in the past with drives that may not be under the vendor's warranty, or if you just want to make sure that what you have is in top notch shape, I would strongly suggest you investigate Spinrite. I consider it an indispensable part of my toolbox. (May, 2005 update: Six years after this article was written I'm still using SpinRite. See my review of SpinRite 6.0 here.)

4. The Importance of Properly Shutting Down

This system works well--as long as files are closed properly, allowing the link information to be recorded. When programs are terminated abnormally, either because of a power failure or system crash, the link information can become lost or damaged. If a file was written to a specific cluster and this information never recorded in the FAT or directory entry, a lost cluster situation occurs.

Such programs as Scandisk can check for this condition and repair the problem by either creating a file from the incomplete link information, or erasing the information and returning the clusters to an unused state. When I run Scandisk and finds lost clusters, I usually direct the program to delete the files it finds because they are not complete. If I directed Scandisk to save the file, it would create a filed named Filexxxx.chk, where xxxx is a number starting at 0000 and increasing by 1 for each chain of lost clusters it finds. If you ever see files like this on your hard drive, this is where they came from.

Double Back-Ups

The designers of the FAT system realized there might be a problem if the area where the FAT was stored became corrupt, so they designed a built-in backup. Two copies of the FAT are stored and in case of dire problems, the backup copy can be used. Stay tuned: next month, I'll tell you how to use the backup copy. Also, look for reviews of other tools that can help when "things go bump in the drive."

Until next month, best in computing.

Part 2

Back to article index

The History of FAT

In order to understand how file problems are repaired, a little history about how DOS/Windows stores its files is in order. The common term for how data is stored on a hard drive is FAT. No, this is not a dietetic word. FAT stands for File Allocation Table. The hard drive's partition(s) is/are divided into what are called allocation units. The allocation unit is the smallest amount of space a file can occupy.

This design allows the data to occupy larger or smaller areas within a single file system. As the size of the partition grows, the size of the allocation unit grows. And, depending on the size of the partition, a file containing a single character can occupy a single cluster of 32KB. There is a limit to the amount of space a FAT (more commonly referred to as FAT16) partition can hold. This limit is approximately 2.1GB.

To support larger hard drives, FAT32 was introduced in Windows 95 OEM Service Release 2. It supports partitions up to 2,048GB. If you keep partitions to less than 8GB, the cluster size is a modest 4K. You pay in compatibility for FAT32, however, as it works in Windows 95/Windows 98 only. You can't access a FAT32 disk from DOS, Windows 3.x, or Windows NT. There is also a slight speed penalty with FAT32 disks, which are approximately five percent slower than comparable FAT16 disks.

Little Clusters of FAT

Clusters (another name for allocation units) are numbered sequentially, starting at zero and ranging to approximately 64000. The system always references partition space by cluster number--never by the physical location of the file.

The directory entry for a file contains the cluster number where the file starts. If a file only occupies one cluster, the FAT entry for that cluster contains a value of zero, indicating the end of the chain. Otherwise, the FAT entry contains the number of the next cluster the file occupies. This forms a one-way chain, starting at the directory entry and continuing to the end of the file.